minimum standards for GC-MS analysis

[vecktor]

A lot of the GC-MS analyses that are being posted online are totally inadequate and unprofessional and in some cases verging on useless.This is bad science. if you are going to go to the effort of running GC-MS then at least do it properly.

A crappy scan of a printout of a non subtracted spectrum with no TIC is not good enough.

posting crap data in the hope that some of the skilled people here will spend their time examining it is disrespectful and unhelpful. the old saying rubbish in rubbish out applies here.

GC-MS data should have the following data as a minimum.

Method:

1, The sample preparation technique whether any derivitization was used
2, The type of injection eg split (with split ratio) splitless on column PTV etc
3, The injector oven temperatures and carrier type and gas flow
4, The Mass spectrometer parameters, in particular the low mass cut off and scan range

Results:
1, The Total Ion Count TIC for the entire run
2, Any selected spectrum should be background subtracted and the background level selected just prior to the peak of interest, in Chemstation this is called the BSB
3, The TIC should be integrated and the integrated results provided to indicate area under each peak.


highly desirable is an extracted ion chromatogram for the 4 largest peaks for the peak of interest or the peak purity should be given, the shape of these extracted ion chromatograms gives a good indication whether these ions are associated with the peak, they also reveal post injector decomposition.

Instead of doing all the data processing on the capture system it is better to supply the raw data. If you are paying for analysis you should expect all the data to be supplied.

for example with chemstation this is the *.d folder with masslynx the *.raw and with finnigan the *.RAW should be supplied etc
the serial number of the MS can be easily removed from the data (in reality as the serial number can be spoofed nobody actually cares)

by supplying the full data then proper processing of the data becomes possible by third parties.

There is plenty of free software to deal with the data.

AMDIS http://chemdata.nist.gov/mass-spc/amdis/downloads/ is a very powerful free tool, you can thank the US chemical weapons verification team for it, for data analysis and can import the following file types,

* Agilent ChemStation (*.D)
* Agilent MS Engine (*.MS)
* Bruker (.MSF)
* Finnigan GCQ (*.MS)
* Finnigan INCON (*.MI)
* Finnigan ITDS (*.DAT)
* HP ChemStation (*.D)
* HP MS Engine (*.MS)
* INFICON GCMS (*.acq)
* JEOL/Shrader (*.lrp)
* Kratos Mach3 (*.run)
* MassLynx NT (*.*)
* Micromass (*.)
* mzXML/mzData (*.*) (new in NIST11)
* NetCDF (*.CDF)
* PE Turbo Mass (*.raw)
* Shimadzu MS (*.R##)
* Shrader/GCMate (*.lrp)
* Varian MS (*.MS)
* Varian SMS (*.sms)
* Varian XMS (*.xms)
* Xcalibur Raw (*.raw)

Wsearch32 is a good freeware program http://www.wsearch.com.au/wsearch32/wsearch32.htm

* Kratos DS90 , AEI DS55
* Kratos Mach3 (*.run)
* Agilent Chemstation (data.ms)
* Agilent LC/MS (msd1.ms)
* HP Pascal. (*.d)
* Varian Saturn (*.ms, *.sms)
* Finnigan ITS40,magnum (*.ms)
* Finnigan ICIS, ITDS (*.dat)
* Finnigan Xcalibur (*.raw)
* Shrader (*.lrp)
* JCAMP_DX (*.dx, *.jdx)
* netCDF (*.cdf)
* Joel (*.lrp)
* mzXML (*.xml, *.mzxml)
* Masslynx (*.raw)
* PE TurboMass (*.raw)
* Bruker XMASS
* Shimadzu (*.R##)


Thanks for listening.




Basic Good Laboratory Practice

it is good laboratory practice to run blanks (solvent only with no sample) before and after each sample, it is also good practice where references are available to run the reference immediately after the sample using the same conditions. The retention time should match within a 0.5 to 1 second. it is also good practice to run a standard mixture regularly. septa should be changed after no more than 200 injections, injector liners should be regularly changed, as should split vent filters, where dirty samples are used or excessive amounts of solvent is injected then the split vent lines and filters need regular cleaning

samples should not overload the column but should give peak heights at least 5 times noise, for example on a clean well set up HP 6890/5973 with a low bleed the background count is usually around 5-8000 counts peak heights should be 20 000 counts at a minimum. Sample injections should never exceed the volume of the injector liner when vaporised except in special cases. low MW solvents like methanol are the biggest offenders here.

samples should dissolve fully in the solvent, in the case of the sample not dissolving then this needs to be noted and attention drawn to this fact. where samples are filtered care needs to be taken to ensure analyte is not absorbed on the filter.

An Internal standard added as a spike to the sample are useful to ensure the sample is getting into the machine, where possible the internal standard should be of a known concentration and should have similar chemical properties to the sample to account for absorption on the injector and column, for example dichlorobenzylamine is a useful IS for phenethylamine type drugs. Naphthalene is quite good too but is inactive so doesn't reveal an active inlet.

The Mass spectrometer should be tuned at least weekly and quick tuned daily. The reduction in heavier mass abundance (502 with PFTBA) and peak shape indicates source clean is required. An air and water check should be regularly performed any increase in N2 and O2 should be investigated.

 

[negrogesic]

For compounds I want to be tested "no questions asked", I rely on the university's postgrads, who use high-end HPLC units. I do not know how to use these devices. I trust these postgrads (medicinal chemistry and MCB, namely), and I've never questioned an analysis. They are also very helpful in respect to data interpretation, but again, especially when there are a variety of compounds in a sample.....

Out of fear of loosing their favour I don't think I will lecture them about proper procedure, but informative post nonetheless. I agree that there are a ton of 'dirty' analytical data floating around, especially in the" grey" fringes...

 

[Le_Messia]

The information is very much appreciated. Great post

 

[mad_scientist]

lol we have had GCMS scans run by the government drug lab before (the same one the police use, and submit as evidence in criminal cases) and they were nowhere near those minimum standards...really makes you wonder how much supposedly damning evidence could be struck out as inconclusive if defense lawyers had any real idea of what those scans mean and what can be proved from them...

 

[vecktor]

lol we have had GCMS scans run by the government drug lab before (the same one the police use, and submit as evidence in criminal cases) and they were nowhere near those minimum standards...really makes you wonder how much supposedly damning evidence could be struck out as inconclusive if defense lawyers had any real idea of what those scans mean and what can be proved from them...

indeed !
The presented standard of forensic analysis is often very poor indeed, and the possibility of carry over between samples is very high, these labs routinely use controlled drugs which end up in the vent line or crashed on the column which are then extracted back out if the solvent volume is too high, or a strong solvent like DCM is used causing the appearance of something that wasn't in the sample (unlucky for the defendant), and don't get me started on the continuity of evidence and fraudulent evidence:-
It is widely known that a certain contract analytical lab was presenting LC-MS based evidence even though the machine was dead and had been for several months (apparently a critical component was with a service company!) and they were just presenting earlier data again and again, just changing the date.
There is also a clear problem with the use of automated library searches too, there have been recent examples of mephedrone being identified as methamphetamine using GC-MS, it has a good library match using some of the algorithms especially if the low mass cutoff is set too high, of course the GC retention time is wrong and doesn't match a standard of methamphetamine but labs don't always run standards rather they rely on the spectrum match. (using a library spectrum match alone is not acceptable, it should also match the retention time and spectrum of a known authentic reference on the same machine)

the prosecution forensic science has basically pitched GC-MS as a gold standard analytical technique: what it says is unquestionable, and the juries buy it. It is very CSI and CSI never gets anything wrong. But any decent analyst is fully aware of the shortcomings of each analytical technique and would never go as far as the government forensic chemists go in their statements, (FWIW it has been my experience that often government analysts and forensic chemists are usually rather mediocre unquestioning people who do not possess the intelligence and aptitude to make it elsewhere.) It is also difficult to find forensic chemists willing to question and challenge other forensic chemists scientifically. It is a small world where everybody knows each other and a lot of careers rely on not pissing off this old boys network, The additional problem is the CSI problem, that juries are just not smart/well educated enough to understand the science so they just buy it all. this is only going to get worse as cost pressures are meaning that sample throughput is the only consideration with forensic analysis, especially with the outsourcing of forensic analysis, the more samples the more money the contractor gets, so blanks and calibration get thrown out, runs are truncated without fully clearing the column, etc etc.
Science is regularly misrepresented and abused in court but they get away with it.

Pharma has had to deal with a lot of these shortcomings and the analytical depts in big pharma companies are usually producing decent valid data. if something doesn't fit it is investigated rather than ignored as it is in the forensic science community. the problem is that there are insufficient samples they do not take or investigate enough samples to ensure that crap doesn't get through, instead relying on manufacturers data and random sampling.

 

[Wizzle]

totally off topic but this mechanism reminds me of issues in the financial crisis. People don't really get pissed because they don't know what's going on. People can lie to you, but even when someone explains you don't know who's speaking the truth. Same goes for things like the Israeli/Palestine conflict.

I don't like to have opinions on these things at all because I trust no single source. Life sucks that way.

I gotta say I AM kind of shocked by this. I have no clue as to how all this stuff works. A friend of mine who works in a lab doing fluid identification and quantification did tell me they used real mdma as a reference. So at least they used their own references. I also know that defendents in Holland often get acquitted because the samples haven't got tested. This is something that happens a lot with minute amounts of controlled substances.

 

[Alcyone]

By all means an excellent thread, but may I suggest that you (we?) make a distinction between strictly qualtitative procedures (identification of the main component(s) and any major impurities) and quantitative procedures (determination of the purity)?

I agree that there is no such thing as too much documentation, but if these are minimum requirements, they are very strict indeed.

I wouldn't like to defend an analytical result in court unless the documentation package included all those data and then some, but for an online evaluation of GC-MS results, I think we could get by with somewhat lower standards. In my personal opinion, sensible minimum requirements for a qualitative analysis would be a TIC, extracted ion chromatograms (traces) of 3-4 characteristic ions for the compound of interest, and of course background subtracted mass spectra of all peaks. Additional information, such as injector temperature or GC column data (make, dimensions, film thickness, stationary phase) should be available upon request. Doing a GC-MS only to present a mass spectrum is blatantly wasteful; essentially they're throwing away at least half of the results.

Quantitative analysis, or true purity determination, is a different matter. A common practice, and a very dubious one, is to run a GC-MS in scan mode, integrate the TIC, and present the purity as the area of the peak of interest relative to the sum of all peak areas. It gives you a ballpark figure at best, and it doesn't take into account the presence of water or residual solvents in the material (invisible by GC-MS) and the widely different chromatographic and mass spectrometric response of different compounds. Impurities present at low concentrations are lost in the background noise, as are compounds with poor chromatographic properties (typically carboxylic acids, phenols, to some extent amines and alcohols). HPLC with UV detection is also frequently used for purity determination. If used properly, HPLC is fine, but if used in this manner it is absolutely worthless. HPLC-UV for identification is beyond worthless, and nobody should be paying money to have a substance "identified" by HPLC.

A proper purity determination by GC-MS or HPLC should rely on a calibration curve, or at the very least a single point calibration, using a certified reference material of known purity, and a suitable internal standard. In addition to the listed requirements, the documentation should include an uncertainty estimate. Measurement uncertainty is often overlooked, but one should keep in mind that even when done properly by experienced people, GC-MS measurements are not exact values. In most cases, anything below 10% (relative) is good, below 5% is extraordinary.

 

[negrogesic]

I seem to get clearer results via HPLC........but again, it is done by others, and they do it as a favor. In regards to identification of unknown substances in a clinical setting, said compound is generally sent out for GC-MS analysis. In some cases compounds are sent to the NIDA for identification, but i've personally only seen this done one time (patient, presented inebriated, combative etc, had a tin foil packet of what smelled like PCP, but tox came up clean; evidently this substance had been found on patients before - I do not know the results of the sample send in, some powerful dissociative or possible a 'qualone'). Etaqualone for example, is not detected on a comp tox screen.

 

[Turing Machine]

I have a question for Vecktor or anyone else who has anything to offer on the subject. In the medical cannabis industry where I live, there's a ton of dispensaries and private labs doing quantitative analysis for cannabinoids. However, it seems obvious that none of them have reference standards for any of the cannabinoids since they have no DEA permits to purchase them and do not have the resources to produce them in house. I've also heard from a friend who is involved with a well known pharmaceutical cannabis preparation in europe that even if they did have the DEA permits, the THC sold by Aldrich and the like are to my surprise, often substantially lower in purity than stated in the catalog. It seems to me like these two factors, combined with a lack of standardization in the protocol for cannabinoic extraction and analysis are the reasons why quantitative testing seems all over the place with many results producing both overall cannabinoid levels which seem to be beyond what the plant should be capable of, and ratios of THC to CBD which defy the genetic models for THC and CBD production. Of course, the dispensaries and labs could just be lying about the results which seems at least equally possible.

Roughly how much could either of these factors alter the analytical results? How accurate can a test be without any sample of THC for reference? I'm considering producing my own reference standard by taking a hexane extract of high purity mechanically separated resin glands from a variety known to be very low in accessory cannabinoids and throw every THC isolation technique I have at it, fractionating under .01 torr, then dissolve in methanol and treat with sonication and potassium hydroxide and filter out precipitate. I presume it would be fairly wasteful but would produce a sample +- 95% pure.

 

[vecktor]

certainly the commercial available reference standards for THC are not pure, as we know THC degrades on storage even when cold, mostly due to it not forming a nice crystalline form it is a glass or very viscous oil and generally oils and glasses are more unstable than solid materials so the reference sample which could be rather old might no longer have the stated amount of material in it.

with relatively little effort it is possible to make material that assays at over 105% purity when compared to commercial references.

I suggest that if you want to make a reference compound that is pretty much absolutely pure THC then you look at the work Mallinckrodt did a few years ago published as US PATENT 7402686, this patent teaches the use of crystalline sulfonate ester derivatives of THC which of course can be purified by recrystalization, avoiding a lot of the hassle of chromatography. I also know GW Pharma developed methods them by which pure THC is extracted from plant sources and again purified by way of crystalline derivatives I cannot find the patent but it is out there, ( I will track it down and post the number later). (IMHO none of these patents are watertight as there are I believe there are prior art references to the use of crystalline sulfonate esters of thc for identification/characterisation going back to before the 1970's Benzenesulphoyl chloride and triethylamine for example ). I would also think that the fair use provisions would apply provided the reference was used internally and wasn't sold. the other huge advantage is that the crystalline derivatives are stable indefinitely at -18 oC.

accurate qualitiative THC measurement is rather difficult without a reliable known reference because THC sticks to glass and to plastic and quite often gets stuck to the sides of pipettes and sample vials. All glassware needs to be silanized to reduce this, the same happens in injectors and columns in GC-MS when underivitised THC is analysed, even so called inert injectors and columns don't stay very inert for long. if you have a reference then the assumption can be made that about the same amount of the reference THC will be lost stuck on the glassware etc so it is fairly simple to calibrate using a series of injections of reference material.

It is better practice to use a derivatising agent to make the THC less sticky, The use of BSTFA or similar silanizing agent to derivitize the cannabinoids is well documented and with care can be done in situ in the sample vial itself. injecting small quantities of BSTFA or similar silanising agent into a GCMS does no harm other than meaning the mass spec source has to be cleaned a little more often.

GC-MS quantification suffers from another problem as Alycone above mentioned, say you have a sample with two compounds present in the same molar concentration, or both compounds are about the same mass and present in equal weights....

one compound A, this is ionised and produces one fragment M+1 or whatever
compound B is ionised and fragments into 4 medium sized fragments

you analyse the sample and then integrate the total ion count TIC, even though the compounds are present in the same molar concentration and there is the same amount of each the TIC will indicate that there is 4 times more B than A, because each molecule of B that reaches the detector makes 4 ions whereas each molecule of A only makes 1 ion.

there are a whole load of other issues and I have barely scratched the surface.

with LC-MS the whole thing becomes even more complicated with ion suppression and other effects that are only minor players in GC-MS.

In the pharmaceutical industry GC-FID and LC with ELSD for heat labile compounds are often used for quantification and assay. FID Flame Ionisation detectors are used for quantifying things because most molecules of a similar mass have a very similar detector response with FID, other than a few chlorinated compounds FID gives a good response that is closely correlated to the number of carbons present.
ELSD Evaporative Light Scattering Detector is a pretty universal detector for LC, which unlike UV or MS sees most compounds, except volatile compounds Bp <100.

one other technique which is very good for quantification where the material is fairly pure is measuring the absorbance (at the maximal absorbance for the compound) of a diluted solution with a UV-VIS spectrometer, and then using the known absorbtion of a reference sample to calculate the concentration in the sample. where the compound does not absorb light strongly a colour reaction can be used to make a coloured derivative product and then measure the concentration of the coloured derivative.

I have to say that I don't know have a huge amount of hands on experience regarding analysing cannabinoids I have only limited practical experience in the area, but I can see there are quite a few potential pitfalls, most of which would be addressed by having a known purity reference for the compounds of interest.

PM me if you want to discuss this further.

V

I have a question for Vecktor or anyone else who has anything to offer on the subject. In the medical cannabis industry where I live, there's a ton of dispensaries and private labs doing quantitative analysis for cannabinoids. However, it seems obvious that none of them have reference standards for any of the cannabinoids since they have no DEA permits to purchase them and do not have the resources to produce them in house. I've also heard from a friend who is involved with a well known pharmaceutical cannabis preparation in europe that even if they did have the DEA permits, the THC sold by Aldrich and the like are to my surprise, often substantially lower in purity than stated in the catalog. It seems to me like these two factors, combined with a lack of standardization in the protocol for cannabinoic extraction and analysis are the reasons why quantitative testing seems all over the place with many results producing both overall cannabinoid levels which seem to be beyond what the plant should be capable of, and ratios of THC to CBD which defy the genetic models for THC and CBD production. Of course, the dispensaries and labs could just be lying about the results which seems at least equally possible.

Roughly how much could either of these factors alter the analytical results? How accurate can a test be without any sample of THC for reference? I'm considering producing my own reference standard by taking a hexane extract of high purity mechanically separated resin glands from a variety known to be very low in accessory cannabinoids and throw every THC isolation technique I have at it, fractionating under .01 torr, then dissolve in methanol and treat with sonication and potassium hydroxide and filter out precipitate. I presume it would be fairly wasteful but would produce a sample +- 95% pure.